SpaceXAI Exodus, OpenAI’s Apple Partnership Sours, iPhone Engineer on Apple’s Roadmap & Steve Jobs

0:12Welcome, everyone, to The Information's TITV. My name is Akash Pasricha. It is Friday, May 15th. First up today, The Information published exclusive reporting about a talent exodus from SpaceX AI. Our Elon Musk reporter will join the show to share with us what he knows. Plus, we'll get our daily update from Rocket Drew about the Elon Musk OpenAI trial with jury heading into deliberation mode. In this week's edition of The Editor's Cut, we're going to dive into the eagerly awaited Cerebrus IPO and a brewing conflict between OpenAI and Apple. Speaking of Apple, we're bringing on an early iPhone engineer to talk about his thoughts on the current state of the company.

0:55And we're going to wrap the show with this weekend's Big Read, spotlighting a nuclear power startup that has set an aggressive deadline to turn on its first nuclear reactor. It's going to be a fun show, so let's get right on into it. XAI is losing talent on a significant scale to other big AI labs. That is according to exclusive reporting from The Information. I want to bring on Theo Waite, our Elon Musk reporter, to walk us through what he knows. Theo, welcome back to the show. it's great to have you here walk us through what you found in your reporting yeah so you know first

1:31Theo Wayt:of all uh we're not allowed to call it xai anymore i noticed in the intro you called it spacexai which is now the official name it's official spacexai you know it could be spacexai kersher for all we know right i mean if you type in spacexai.com right now it doesn't actually go to anything so they still need to figure that out but at least according to elon it's what we're supposed to call it okay um so anyways spacex ai uh it's a mouthful uh there's there's been sorry i just got i i gotta jump is the x lowercase now or is it uppercase it's friday so we gotta square it away because spacex it was a capital x right i think it's uppercase officially elon says it's called spacex ai the ai products from spacex so try saying that five times fast um it's all right all right so what did you find about the people working at spacex ai the company formerly known as xai and so since spacex acquired xai in february uh just to keep the timeline of the names straight uh there have been at least 50 research staff who have left more than 50 um and that's pretty significant because late last year there were uh somewhat somewhere in the range of like over 200 but you know 50 of those that's that's a pretty big proportion um and that doesn't include other departures like the cfo who left a bunch of people working on data centers who left ai training tutoring staff who also left this is just researchers so it's a it's a pretty big number that i was able to identify from from reporting um and then you know in terms of where they're going, Meta has been a big destination.

3:19Theo Wayt:In the story, I said at least 11. Since it's published, I've already increased my tally, so it's even more than 11 at this point. At least seven are going to Thinking Machines Lab, Miramirati's company. And then there's a smaller number going to other companies like Anthropic and others. But Meta and Thinking Machines are definitely the biggest two destinations. So this is kind of interesting. This is kind of a look into how reporting works behind the scenes. But since you've published, you've actually identified even more people who have left, you're saying? Yeah, I mean, you know, I would just say if you're someone that's considering messaging me on Signal, please do it before I put the story out instead of after.

4:01Theo Wayt:But yes, I, you know, I've been keeping this like growing list of people that have left and it just keeps growing even when I already published the story. so now you talked about them going to meta and to other companies and in the story you talked a little bit about the reasons why that might be the case you identified the culture at the company being one factor i mean what do we know here about it was this you know meta poaching uh people was it other companies poaching people or was it people disgruntled saying i want to leave uh i think all of the above i mean to be clear like a lot of these departures are um either firings or layoffs i mean there have been quite a few of those it's it's not all people choosing to leave um you know there are plenty of people that were gung-ho about working at xai or sorry space xai that um that that you know got pushed out regardless um the other factor for a lot of people, I think, is just that the company has this six-month period after which you vest some stock.

5:09Theo Wayt:And so a lot of people are just trying to make it to six months, and a lot of them don't. I mean, a lot of the people that have been let go or left recently were there for even less time in six months. Musk has said that he did not build XAI right the first time. This is something that we've talked about on the show what do we know about how he has changed the way that spacex ai now runs tactically on the ground i mean it kind of seems like he's do like he's kind of doing a version of doge to his own company almost uh you know that might be overstating it a little bit but there are some similarities like he he you know has this team of people from spacex that he trusts um and this team of people from cursor that he's you know now has this this arrangement with and they're kind of going into xai and you know or what was formerly known as xai and going through and looking for places to make changes people to cut um and you know fat to trim etc um which you know is different from doge in that they're doing it to a company that elon himself started So it's, you know, kind of the snake eating its own tail or whatever.

6:28Theo Wayt:But there are similarities here where like, you know, Elon gets really focused on something and he has to, you know, start cutting and making heads roll and blaming people for things. And that's what's going on right now. What about Kurser? You had some reporting last week, I believe it was, about the current state of Kurser and how that company is dealing with this new deal that conceivably is going to come together with SpaceX AI. What's going on over there? Yeah, so Cursor staff have been coming into the XAI offices and talking to people about what they're doing. That started happening around the same time there were some layoffs.

7:13Theo Wayt:So a lot of people at XAI that I've talked to kind of see a connection between those two, although I think cursor would say that it's more just like getting to know you meetings. And Elon has also said that XAI is using cursor data to improve its coding models. So there's a lot of movement in that relationship, but I think it's still pretty unclear whether SpaceX is actually going to go through with this option that they have to buy cursor. And it'll be really interesting to see what the SpaceX S1 that's supposed to flip as soon is next week is going to say about that because it's kind of a strange arrangement.

7:53Theo Wayt:They haven't really detailed very much at all. So this culture that you describe here of a bit of a revolving door, people coming in and out, I wonder if you think that is going to change at all as SpaceX AI becomes a public company. And I ask this because we have Tesla, which has been a public company for a long time. Do we have any reason to believe things will change when they're, you know, under the scrutiny of public shareholders? Or is it really just going to be the same? I mean, I think it's an Elon Musk company more than anything. And, you know, all signs point to him having even greater control over SpaceX than he does with Tesla.

8:36Theo Wayt:So I wouldn't expect it to change a whole lot. I mean, it seems like, you know, with any of his companies, he gets very focused on an area. It gets very intense. He fires people, people quit, and then he moves on to another project and maybe it's more stable for a while. And that's happened a lot at XAI recently, but it's also happened to Tesla during times that that's been a public company. It's just how he works. Great. Well, Theo, I want to thank you for coming on. That is Theo Waite, our Elon Musk reporter here at The Information. The Information's Rocket Drew is covering the Elon Musk OpenAI trial from the courtroom all week long.

9:15Here is his latest update. It was the longest day yet in the trial for Elon Musk's lawsuit against OpenAI and Microsoft, and that's because it was the day for each side to deliver their closing statements. That's the opportunity for each side to go through the mountains of evidence that have accumulated during the trial, pick out the pieces that they think are most important, and stitch them together into a story that hopefully will be compelling to the jury and convince the jury to side with them. That's really a service to the jury at this point in the trial, because there's so much evidence that's piled up.

9:52There's been email threads and text exchanges and all sorts of documents from across the history of OpenAI involving people both inside OpenAI and outside. So for the jury to be able to know where to begin and even make sense of these documents, it's really helpful for the lawyers to provide them a sort of legal framework. Now, at this stage in the lawsuit, Elon Musk has two claims against OpenAI. That's breach of charitable trust and unjust enrichment. Lawyers for both Elon and OpenAI spent most of their time on the breach of charitable trust claim. To show that there was a breach of charitable trust seems kind of straightforward.

10:31You have to show that trust was created and that the trust was breached, meaning some kind of promise or obligation was not followed through on. But after listening to both sides, I really got the impression that there's something structural about the law that favors OpenAI's side of the case here, which is that Musk needs to show all of the elements for breach of charitable trust, and if OpenAI can disprove even a single one of them, that's sufficient for OpenAI to win. So in order to create the charitable trust in the first place, there has to be a donation, and both sides do agree that Musk donated around$38 million and also some non-monetary contributions.

11:11But things get complicated from there because Musk's side needs to show that those contributions were given with a specific charitable purpose. And OpenAI says there was no charitable purpose beyond just supporting the mission of OpenAI, and that's not specific enough. Musk's side says that's plenty specific because Musk had a hand in crafting the OpenAI Charter and figuring out what its mission was going to be in the first place so the charter should count. So was there a specific purpose? Second is whether that purpose, whether there was a proper manifestation of that purpose, meaning Musk made clear, he communicated that there was some enforceable obligation here that he intended to enforce.

11:52And third, that the trust property, in this case the donations, still existed at the time of the breach. And that's again something that OpenAI contests. They say money was coming in the door from Musk, but it was also leaving. They were spending it very quickly because they're a startup, they're spending a lot of money on compute, on talent. So by the time any alleged breach did happen, they didn't have any of Musk's money in the bank anymore. Musk's lawyers counter that it really is sensitive to how you perform the analysis, and a lot rests on the assumption that when Musk's dollar is the first dollar into the bank account, that it will also be the first dollar spent.

12:31And without that assumption, the timeline becomes less clear. They also argue that the money was then spent on assets and employees that were still with OpenAI at the time of the alleged breaches. So there was some back and forth. There was some clash over that point. Microsoft is in an even better position with respect to this claim because the claim against Microsoft is aiding and abetting breach of charitable trust. It's derivative of the claim against OpenAI, meaning the jury will only reach the question about what happens to Microsoft if they already agree that there was some breach of charitable trust that OpenAI committed.

13:06If not, Microsoft is off the hook. But supposing they do get to Microsoft, Microsoft still has a number of outs available to them. They can say that they didn't have any knowledge of the charitable trust that was formed with Musk. They weren't aware of any commitments that were tied to his contributions, and they say that's the case. They did their due diligence, and nothing about this ever came up before they invested in OpenAI. And Microsoft can also say that they were not a substantial factor in the breach, and they were not a substantial factor in any harm that was caused to Musk. So that puts them in an even safer position in this lawsuit.

13:45The last thing is that both sides, both the defendants, OpenAI and Microsoft, have some further defenses. They argue that the statute of limitations on the claims that Musk is bringing have expired, meaning he waited too long to bring the lawsuit. Whatever his complaints were, he should have known about them originally. He should have known about them, say, in 2020 when he started tweeting at OpenAI and Microsoft that it seemed like they were heading in a direction that was not in line with his original expectations. uh musk's side counters that they actually didn't know that the breach had occurred until 2022 in fact when musk saw an article from the information outlining the deal talks that were happening between opening eye and microsoft at the time musk's lawyer says that was the tipping point for him that was the breach and there were some unique features of that 2023 deal that made it so that was the violation whereas previous deals with microsoft did not meet that threshold for one the The volume of the investment was much larger.

14:43It was a$10 billion investment, which is greater than the$3 billion Microsoft had invested before. And the 2023 investment also involved a change to the profit caps that were implemented for the previous investments that OpenAI had received. In 2023, the parties agreed that those profit caps would increase year on year by 20%. So every year that goes by, the profit caps get 20 % higher. That's that much more money. That's that much more profits that would have to be returned to OpenAI investors before they sort of hit their caps and then the residual begins flowing to the nonprofit. So that was another feature of the 2023 deal that Musk's side objected to.

15:28Musk's side also, they critiqued the credibility of the witnesses on OpenAI's side, both Sam Altman, they spent a lot of time critiquing Sam Altman's trustworthiness and credibility when he was on the stand, but also Greg Brockman. They've pointed out that Greg Brockman has about$30 billion worth of equity in OpenAI, and that means that he obviously is interested, very personally interested in the outcome of the lawsuit. And they also suggested that he otherwise was not a reliable witness while he was on the stand. So it was a long day. There were some heated back and forth. There were some really interesting points of contention, but now the ball is in the jury's court.

16:08So beginning Monday, the jury will start deliberating in earnest. At the same time, the judge will begin hearing the next phase of this trial while the jury is not in the room. That's the remedies phase. So supposing the jury comes back and has a finding of liability and the judge agrees with that finding, which she may not, but if she does, then there will be a question of what remedies the court should grant. So the judge is going to get started right away on hearing the remedies portion of the trial. That will involve some testimony from expert witnesses, economists who have things to say about what sort of remedies would be appropriate in a case like this if there is a breach of charitable trust or there is unjust enrichment.

16:46But that's more to look forward to next week in Musk v. Altman.

17:16San Francisco Bureau Chief for this week's edition of the Editor's Cut. Welcome to you both. Martin, I want to start with you. So is OpenAI an Apple draw? Did we see this one coming? No. In fact, Elon Musk had filed legal action against both of them, claiming that they were colluding against him. So I don't think anybody saw this coming. But, you know, I think we should seen it coming because you remember that not that long ago openai acquired tbpn putting it into the tech media business and i really think that now we understand why they did that openai is one of the greatest uh sources of tech news that we have seen in a long time uh aside for only freely from elon

18:08and in that regard, owning a tech media outlet means that they are vertically integrated. And so this is just the latest example of that. I mean, really, it's a sign that this company is not very mature and they need a real PR team because what company leaks so obviously? All the reports were quoting an unnamed OpenAI executive complaining about Apple. It just seems like amateur hour. Okay, so Jason, what do you think? I'm not going to come back to you, but Jason, what are your thoughts? I think there's probably more to it than trying to juice CDPN's ratings, but that is a fair point by Martin.

18:58I think, look, Apple's failure to do anything impressive with AI has to be disappointing to OpenAI. I mean, in a way that's a little bit hard to imagine, they actually lent Apple quite a bit of credibility when they, you know, AI cred when they did that deal. And, you know, Apple's repaid that by burying this open AI interactivity in the bowels of their labyrinthine settings system. You really can't find it unless you, like, do a chat GPT search about where to go to turn it on. It's been profoundly disappointing. Now, it's hard to imagine that Apple is failing at AI just to spite ChatGPT, which almost seems to be the subtext of their complaints.

19:55But, you know, one can imagine why they would not be pleased with how this has panned out. And obviously, Apple is now linked up with Google and Gemini and may well be announcing other AI partners at WWDC Developer Conference next month. I have to say, I don't agree with that take at all. To me, chat GPT has become the byword for AI. I mean, it's a brand name that everybody knows. The fact that OpenAI is complaining about Apple, it's almost like they're blaming Apple for the issues that they've had with their business, which is really their own fault and their complete lack of any kind of focus.

20:42I mean, I really just don't understand. I mean, they don't need Apple to be successful. And I don't really see that, you know, the fact that Apple hasn't done so well in AI has anything to do with or has any impact on open AI. You know, I think it's just their own entirely open AI's fault that they have screwed things up because they've been so focused on so many different things. Well, and I just want to backtrack a second here, Martin. I mean, let's be clear here. What is the substance of the, you know, complaint that they're considering as far as we know? I mean, Jason sounds like he knows, he understands it more than I do.

21:22I think it is essentially that they haven't gotten out of the partnership what they were hoping to get. But as I said, I just, to me, that's, I mean, what were they expecting? Well, Jason, I mean, Jason, let's - on that they should not have been relying on iPhone distribution for their own success. Anyway, that's just my view. Well, Jason, what do you think? Because, you know, one question I've had is like the original partnership asked, what was even the substance of the original partnership when they announced? I'm recalling, I think it was a keynote, right? I mean, Tim Cook came out. He said OpenAI will be used in some capacity, right, for Apple.

22:07I think this was the same keynote where they released Apple Intelligence, right, if I'm not mistaken? That's right, yeah. And, you know, it was never supposed to be exclusive. And, you know, I certainly take Martin's points, but somebody's got to argue the opening eyesight of this, and I would say, like, I would say, yes, yes, they've succeeded, but in spite of Apple, and what did they expect to come out of this? Well, they might have expected a company that for a long time was the highest valued in the world and that brags about its innovative abilities to be a little more competent at doing AI and integrating that into its platforms.

22:52I mean, iPhone has been big for ChatGPT, but that's just because people like to use the app. It's not really got anything to do with how Apple has implemented and executed on that original partnership, even though it, yes, was limited from the get-go. Okay, so I want to move to another big headline of the week. So Cerebrus went public. The stock popped. Martin, we have seen this before. What are you expecting in the weeks to come? I'm expecting it will gradually fall. It's already down this morning. the whole market is down and it is falling with it i mean um you know we're in a very upbeat mood right now about about ai i think that that's good for uh the other companies that are planning to go public as long as it lasts now i don't think it will last i think there's a sort of a widespread expectation that the market is very um you know high right now and is going to fall and i think one of the issues about spacex is they're probably trying to get out as fast as they can before the market falls um and you know you know what will happen at that point it will be harder for it for anyone to go um public and there is this overhang of the oil um price and the fact that the war in iran has created there isn't enough oil for the world and that's going affect the economy and once that begins to really flow through to things I think you will see the market will fall and I just it's now just a question of time so I think this is a very short-term blip that we've seen and Cerebra stock which is I think this morning is around 289 will probably be back around 115 which is where they were originally planning to go public.

24:50So Jason, because you love a good debate, are you going to argue now that the market is going to continue to rise? Well, you know, I'm not going to prognosticate steady, unmitigated, up and to the right. I think, you know, markets in the short term can be choppy, but, you know, long-term holders here are thinking about horizon more than the next six to 12 months. And if you're looking at it from that perspective, yeah, the future could be quite bright for Cerebris. I mean, yes, their valuation is like astronomically higher than that of NVIDIA, even despite this morning's dip on a price to earnings multiple.

25:35But that's because they're going to take market share from NVIDIA. The future is inference. It's going to be a much bigger part of AI compute than it has been. It's already pretty huge. And, you know, they have chips that are not only bigger than everybody else's, which, let's face it, is pretty cool, but also are just, like, better at doing inference and solving a lot of the problems that GPUs from NVIDIA and other companies have at that task. So I think, and this is not a Johnny-come-lately either. They've been around for a decade. They've got a pretty proven track record at developing this technology, and they have some pretty important partners.

26:12So, yes, it may not stay aloft at the altitude. It hit yesterday in the near term, but I think longer term, you can make a case that there's a reason to be bullish. Listen, how big do you think they can get? I mean, it seems to me that they're up against an awful lot of competition, and I don't really understand how they can gain. I mean, they're there as an option when people can't get other chips, but really, are they going to be able to defeat NVIDIA long term or even Google or Amazon chips? I don't think so. They don't need to completely defeat those companies to have a massive future business.

27:00I don't know the percentage exactly, but I would say there is a lot of interest from a lot of powerful parties for there to be a much more diverse chip supply ecosystem for AI than there has been. I mean, this industry, many of the inflated costs and problems and bottlenecking all come from the fact that the industry is still absolutely dependent on NVIDIA GPUs. That's clearly not going to be the case in the same way, you know, three, five, ten years from now. And it won't just be Cerebrus, obviously, as you point out. Google and Amazon are building their own chips and making a lot of progress.

27:45but it's going to be a bigger pie and i think there's definitely a chance for cerebrus to have a bigger piece of it or martin one more op you floated this option in your briefing the other night and you know it might be a little harder now that the stock is up so much but i mean cerebrus could get bought potentially and and you know their technology although i think the chance of that have probably gone out the window with the stock going up so much so i don't know why anyone would pay the kind of price that you would have to pay now. So, yeah, I think that was an option before the IPR. And, you know, if the market had not been like this, things might have gone very differently.

28:26So, I guess we'll see. Great. Well, I want to thank you both for coming on. That is Martin Piers, our co-executive editor, and Jason Dean, our San Francisco Bureau Chief, here at The Information. Our next guest is a well-established expert in building products that people will actually use. Matt Rogers was an early engineer of the iPhone at Apple. He is the co-founder of Nest, the thermostat company that sold to Google for$3.2 billion. And now he is working squarely in the sustainability sector with a food waste company called Mill. I want to bring him on to chat about all of his points of view on the current state of play in Silicon Valley.

29:05Matt, welcome to the show. It's great to have you here. Thanks for having me. I'm excited for this one. So look, I mean, let's, can I just understand this? You went from iPhone to thermostat to trash cans. That's the career trajectory. You got, actually folks laughed when Tony and I started Nest the first time around too. Look, it's the, and these are the kinds of problems that I feel like as an engineer and designer, I became an engineer to solve. There's so many people chasing the latest shiny object, but there are just so few world-class designers and engineers working on these really important real world problems yeah yeah so tell me about the tell me about the the trash can it's more than a trash can I'm being I'm being I'm being a little wise here yes so so Mill is a food recycling company it combines hardware and AI to make it easy for people at home and businesses to reduce their food waste learn more about what food is being wasted and to actually save a lot of money uh believe it or not the value of wasted food in America is like$400 billion a year.

30:08It's like one plus percent of GDP. It's like giant problem that we live with every day that we actually could do a lot about. And how did you get hooked on this idea, you know, coming out of the thermostat company? Yeah, I've always been obsessed with these kinds of things. Like part of my training at Apple was to be able to see these things and you know, we do every day and actually think about them. And during the early days of the pandemic, when we were all stuck at home living in our own filth, it became very obvious that this is one of those problems that could be solved. Like my house was getting full of fruit flies and there had to have been a better way.

30:43Yeah, but like how, I mean, just help me understand. Like, you know, what are you, sitting there with fruit flies and you're like, I have, like, what was the moment that you were like, I'm going to build a company to solve this issue? I mean, you could have done anything. Yeah, so like post-Nest acquisition, my wife and I do a lot of investing in climate companies. We do a lot of philanthropy and climate. And waste is a big part of the climate equation. It's like 10 % of global emissions is food waste. So the problem is really well known. Living with the kind of the fruit flies, trying to do the right thing, trying to compost at home, it became very clear like this is not an easy thing to do.

31:23And if it's this hard to do, no one's going to do it. You know, it's kind of like the old adage of like, people are going to find the easiest way out. And if composting at home is hard, if composting at work is hard, just no one's going to do it. So we set out to make it an easy way to do it. And, you know, following our normal kind of engineering and design process, we arrived at what became our first product. And we've obviously made several iterations since then. And I have to ask, because I'm just so fascinated by this product. Like, can you just explain how the composting actually happened i mean so i i put my my leftovers into the into the trash bin what actually happens to compost something you know and then turn it into mill at the end of the day it is straight up magic uh and what happens is you put all your food scraps in you kind of wipe your dinner plate in you're cutting board scraps you could fill it up if you want yeah right pounds of food waste in this thing and you wake up the next morning and it's gone.